subject raising in zulu and the nature of predp

44
DOI 10.1515/tlr-2013-0013 The Linguistic Review 2013; 30(3): 423 – 466 Leston Chandler Buell and Merijn de Dreu Subject raising in Zulu and the nature of PredP Abstract: The Bantu language Zulu has several types of predication in which no verb appears: adjectival and nominal predication, locative and possessive predi- cation, and two types of existential predication. We show that these predication types differ from verbal predication in generally disallowing an internal subject (i.e., an unraised external argument), even in the case of nominal predication, which seemingly displays the unraised subject agreement pattern. Like Zeller (2010, 2011), we assume that the syntactic peculiarities of non-verbal predicates are attributable to the projection of PredP, whose head is morphologically real- ized in nominal predication. However, contra Zeller, we argue that PredP does not also project in certain verbal inversion constructions with internal subjects. The properties of these inversions are argued to be better captured by more general principles, such as the aboutness property of the preverbal subject position. Fur- thermore, only restricted projection of PredP captures the morphological alterna- tion found in certain prefixes which follows the verbal/non-verbal dichotomy. It is argued that the existential predicate -khona, which has properties of both verbal and non-verbal predicates, is a special kind of verb, the only one on top of which PredP can project. Keywords: predication, expletive constructions, subjects, PredP, Zulu Leston Chandler Buell: University of Amsterdam. E-mail: [email protected] Merijn de Dreu: Leiden University. E-mail: [email protected] Brought to you by | Georgetown University Authenticated | 141.161.91.14 Download Date | 9/27/13 3:27 PM

Upload: merijn

Post on 19-Dec-2016

239 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Subject raising in Zulu and the nature of PredP

DOI 10.1515/tlr-2013-0013   The Linguistic Review 2013; 30(3): 423 – 466

Leston Chandler Buell and Merijn de DreuSubject raising in Zulu and the nature of PredP

Abstract: The Bantu language Zulu has several types of predication in which no verb appears: adjectival and nominal predication, locative and possessive predi-cation, and two types of existential predication. We show that these predication types differ from verbal predication in generally disallowing an internal subject (i.e., an unraised external argument), even in the case of nominal predication, which seemingly displays the unraised subject agreement pattern. Like Zeller (2010, 2011), we assume that the syntactic peculiarities of non-verbal predicates are attributable to the projection of PredP, whose head is morphologically real-ized in nominal predication. However, contra Zeller, we argue that PredP does not also project in certain verbal inversion constructions with internal subjects. The properties of these inversions are argued to be better captured by more general principles, such as the aboutness property of the preverbal subject position. Fur-thermore, only restricted projection of PredP captures the morphological alterna-tion found in certain prefixes which follows the verbal/non-verbal dichotomy. It is argued that the existential predicate -khona, which has properties of both verbal and non-verbal predicates, is a special kind of verb, the only one on top of which PredP can project.

Keywords: predication, expletive constructions, subjects, PredP, Zulu

Leston Chandler Buell: University of Amsterdam. E-mail: [email protected] de Dreu: Leiden University. E-mail: [email protected]

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 2: Subject raising in Zulu and the nature of PredP

424   Leston Chandler Buell and Merijn de Dreu

1 Introduction1

Zulu is a Bantu language of the Nguni subgroup, spoken primarily in South Africa and to a large extent mutually intelligible with Xhosa, Swati, and Ndebele.2 While there is a lexical verb at the core of most predicates in Zulu, such as the sentence in (1), there are also numerous types of non-verbal predicates, such as the adjec-tival and possessive predicates illustrated in (2) and (3).

(1) U-malume u-theng-e imoto e-ntsha. det-1a.uncle  1sm-buy-prf.cj  det-9car  9rel-9new ‘My uncle has bought a new car.’

(2) I-moto i-bomvu. det-9car  9sm-red ‘The car is red.’

(3) U-ne-moto e-ntsha. 1sm-with:det-9car  9rel-9new ‘He has a new car.’

This paper is concerned with certain morphological and syntactic differences between verbal predicates on the one hand and the non-verbal ones on the other. On the morphological side, two pre-stem morphemes (the persistive marker and a negation marker) take on distinct forms when they appear with any of the non-verbal predicates. (See also Van der Spuy [2001].) On the syntactic side, while verbal predicates readily allow the logical subject to remain in a vP-internal posi-tion, non-verbal predicates, with one exception, never allow their subject to remain in an analogous low position, but rather always require their subject to raise. We will attribute this morphological and syntactic divide between verbal and non-verbal predication types to PredP, claiming that, in accordance with Baker (2003), but in contrast to Zeller (2012), this category projects itself only

1 In the article, the following abbreviations are used: cj = conjoint, cop = copular particle dj = disjoint, det = determiner, dfut = distant future, epen = epenthetic, fut = (near) future, fs = final suffix, loc = locative, neg = negative, neu = neutral, psv = passive, perf = perfect, pred = predicate, pres = present, pron = pronoun, prs = persistive, pst = (remote) past, rel = relative subject marker, rem = remote, sm = subject marker, sbjv = subjunctive. Numbers refer to noun class unless they are followed by an sg for singular or pl for plural, in which case they denote person.2 For a short general overview of the Bantu S-zone languages, to which the Nguni languages belong, see Gowlett (2003).

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 3: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   425

when the type of predicate embedded below it is non-verbal. Under our analysis, it is sensitivity to the presence of PredP that drives the morphological alternation, while it is absence of head-movement to Pred° that accounts for the obligatory subject raising in non-verbal predication.

2  A short introduction to Zulu predicative morphology

This article focuses on post-predicate subjects in non-verbal predication. To be able to parse the examples, it is instructive to look at verbal predication first. Zulu verbal morphology is complex, and a finite verb follows the following template:

(4)

negativemarker/relativemarker

subjectagreement

negativemarker

TAM morphemes

objectagreement

root extensionsfinalsuffix

The first negative marker is only used for verbs in the indicative principal (unsubordinated) submood, the second for non-indicative verbs and participial (subordinated) submood, which is used in a variety of subordinate clause types, including relative clauses.3 There may be more than one morpheme in the TAM (tense, aspect, and mood) slot. Extensions are morphemes that are used for valence-changing morphemes like passive, causative, and applicative. Similar to the TAM slot, a verb may have multiple extensions. The final suffix varies with tense, mood, and modality, but is often the default -a.The verb agrees with the subject either in person and number or (in 3rd person) in noun class. A verb may also show agreement with its object, in this case also either in person and number or (in 3rd person) in noun class. Subject agreement, the root, and the final suffix slot are obligatory, the other slots are optional. Here are a few examples:

3 Note that participial mood is not the same as a participle. Predicates in participial mood have structures resembling indicative matrix clause predicates in both form and complexity. For example, they both have subject markers and can bear the same range of tense and aspect morphology.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 4: Subject raising in Zulu and the nature of PredP

426   Leston Chandler Buell and Merijn de Dreu

(5) a. A-si-zu-cul-a. neg-1pl.sm-neg.fut-sing-fs ‘We won’t sing.’ (Buell 2005: 121) b. Si-sa-lamb-ile. 1pl.sm-prs-get.hungry-prf.dj ‘We are still hungry.’ (Buell 2005: 112) c. U-moba o-mningi u-tshwal-w-e ku-lesi det-3sugar.cane  3rel-3much  3sm-cultivate-psv-prf.cj  loc-this7 sifundisa. 7province ‘Much sugar cane is cultivated in this province.’

In Zulu, non-verbal predicates can be prefixed with morphemes of any of the pre-root slots of (4) except object agreement.4 Most of the morphemes used for non-verbal predicates are the same as for verbs the negative marker-nga-and the persistive marker -sa- are realized as -nge-and-se-,respectively.

(6) a. i-zingane ezi-nge-zincane det-10children  10rel-neg-10small ‘children who aren’t small’ b. i-zingane ezi-se-zincane det-10children  10rel-prst-10small ‘children who are still small’

The various types of non-verbal predication are only inflected in the indica-tive present tense, although in different submoods (principal and participial), aspects, and polarities. They also lack the alternation between disjoint and con-joint forms, discussed in Section 3. (7a) is an example of nominal predication in the present tense, inflected for first person singular. (7b) and (7c) are also in the present tense, but here they are combined with the persistive marker -se- and the negative marker a-. (7a–c) are all grammatical, because they are in the present tense. In contrast, (8b) illustrates the fact that non-verbal predicates cannot be inflected for tenses other than the present.

4 Certain TAM morphemes cannot appear directly attached to a non-verbal predicate, but some morphemes associated with the slot can.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 5: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   427

(7) a. Ngi-ngu-mfundisi. 1sg.sm-cop-1teacher ‘I am a teacher.’ b. Ngi-se-ngu-mfundisi. 1sg.sm-prst-cop-1teacher ‘I am still a teacher.’ c. A-ngi-yena u-mfundisi. neg-1sg.sm-1pron  det-1teacher ‘I am not a teacher.’

(8) a. Ngi-zo-fundis-a. 1sg.sm-fut-teach-fs ‘I will teach.’ b. *Ngi-zo-ngu-mfundisi. 1sg.sm-fut-cop-1teacher intended: ‘I will be a teacher.’

Not all of our examples are in the present tense. For other tenses the auxiliary be is used in combination with the predicate in the so-called participial submood of the present tense, which amongst other things manifests the form e- for noun class 1 subject agreement, rather than the usual u-. In the example in (9), the verb form uyobe is in the remote future of the principal indicative mood, while the embedded nominal predicate is in the participial indicative of the present.

(9) U-yo-be e-ngu-mfundisi. 1sm-dfut-be  1sm-cop-1teacher ‘He will be a teacher.’

When be is in the recent past or remote past tense, a shorter version of this auxiliary is often used that cliticizes onto the following word, as in (10b) and (11b):5

(10) a. Ng-a-be ngi-ngu-mfundisi. 1s-pst-be  1s-cop-1teacher b. Ng-a-ngi-ngu-mfundisi 1sg.sm-pst-be-cop-1teacher ‘I was a teacher.’

5 Different sources give different paradigms for these shortened versions, and in some cases the participial subject agreement of the predicate may be deleted. See Buell (2005: §4.1) and Poulos and Msimang (1998) for a discussion.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 6: Subject raising in Zulu and the nature of PredP

428   Leston Chandler Buell and Merijn de Dreu

(11) a. Ngi-be ngi-ngu-mfundisi. 1s-be:perf  1s-cop-1teacher b. Be-ngi-ngu-mfundisi. be:perf-1s-cop-1teacher ‘I was a teacher.’

A word is in order over our use of the term “non-verbal predication”. By this term we mean a type of predication in which there is no lexical verb, no V°, a predicate type in which the core of the meaning lies in something other than the verb. The term should not be taken to mean that a verb is necessarily absent, as seen, for example, by the presence of the auxiliary verb -be in the nominal predi-cation examples in (10) and (11). There is even a plausible analysis available, dis-cussed in Section 5, which maintains that an underlying copular verb is present even in cases like (2) and (3), in which no such verb is visible. We consider such auxiliary and copular verbs to belong to a syntactic category other than V and assume them to be introduced in a position higher than V in the structure. A clear-cut distinction between verbal and non-verbal predication can be found in many divergent language families (Hengeveld 1992).

Having given a general description of the morphology of non-verbal predica-tion, we will now discuss the syntax.

3 Right-dislocation in Zulu

Though the unmarked word order in Zulu is SVO, as above in (1), post-nominal subjects are also very common. If a subject is post-verbal, the verb may have two different forms. In one case the verb shows full subject agreement, that is, with person, number, and noun class, while in the other case it has default noun class 17 agreement. In some cases there is an additional difference: when the verb shows full agreement it is in the so-called disjoint form and if it has default noun class 17 agreement it is in the so-called conjoint form, as illustrated in (12).

(12) a. Ku-fik-e i-zingane.] 17sm-arrive-prf.cj  det-10children ‘The children/some children have come.’ b. Zi-fik-ile] i-zingane. 10sm-arrive-prf.dj  det-10children ‘The children have come.’

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 7: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   429

In (12a) the verb shows default noun class 17 agreement, and the perfect -eending shows that the verb is in the conjoint form. In (12b) the verb shows full agreement, in this case with noun class 10, and the perfect -ile ending shows that the verb is in the disjoint form. Clauses with an internal subject and class 17 sub-ject agreement as in (12a) will henceforth be termed “expletive subject clauses”.

Since Van der Spuy (1993) and Buell (2006), it is now standardly assumed for Zulu that the subject of a verb that is in the conjoint form and has default noun class 17 agreement is VP-internal, and the subject of a verb that is in the disjoint form and has full subject agreement is dislocated. We have indicated this in (12) with a closing bracket. Material immediately to the left of the bracket is in some “predicate-internal” position, such as spec-vP, while any material to the right of the bracket is dislocated. We will hereon assume that both the conjoint/disjoint alternation and the use of an agreeing or class 17 subject marker are diagnostics for in situ or dislocated subjects, while we will use both phenomena simultane-ously in our examples when that is possible. Evidence that the two occupy differ-ent positions includes data concerning two different types of phenomena: focus and augmentless nouns. We will now present the these two briefly.

We begin with focus-related phenomena. A vP-internal subject can be fo-cused, while its right-dislocated counterpart cannot. This is most easily illus-trated by modifying the subject with kuphela ‘only’. As shown in (13), only the non-dislocated subject can be modified in this way.

(13) a. Ku-cul-e u-Sipho kuphela.] 17sm-sing-prf.cj  det-1a.Sipho  only ‘Only Sipho sang.’ b. *U-cul-ile] u-Sipho kuphela. 1sm-sing-prf.dj  det-1a.Sipho  only ‘Sipho sang.’

Similar facts can be shown for related focus phenomena such as contrastive focus and in situ constituent questioning.

A noun in Zulu usually has an augment, also called a pre-prefix or initial vowel, which is a sort of determiner (Visser 2008; de Dreu 2008; Buell 2011) and which is used with both definite and indefinite nouns. In our examples we have glossed it as “det”. Augmentless nouns (often called “bare nouns” in the litera-ture) behave as negative polarity items and can appear within the verb phrase of a negative clause headed by a lexical verb (Halpert 2012).6 (14a) and (14b) show

6 Augmentless nouns also have other uses, such as in vocative phrases and after demonstra-tives. See von Staden (1973) for a detailed description.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 8: Subject raising in Zulu and the nature of PredP

430   Leston Chandler Buell and Merijn de Dreu

that in an affirmative clause augmentless nouns are prohibited, while (14c) and (14d) show that augmentless nouns are allowed in a negative clause.

(14) a. Ngi-fun-a i-mali. (with augment) 1sg.sm-want-fs  det-9money ‘I want (some) money.’ b. *Ngi-fun-a  mali. (augmentless) c. A-ngi-fun-i i-mali. (with augment) neg-1sg.sm-want-neg  det-9money       ‘I don’t want money.’ d. A-ngi-fun-i mali. (augmentless) neg-1sg.sm-want-neg  9money ‘I don’t want any money.’

As shown by the postverbal logical subject muntu ‘person’ in (15), an aug-mentless noun cannot be right-dislocated. In (15a) the verb shows default noun class 17 agreement, indicating that the augmentless noun is vP-internal. In (15b) the verb shows full agreement with the class 1 subject marker -ka-, so the post-verbal subject is necessarily dislocated. As the ungrammaticality of the sentence shows, this configuration is impossible for augmentless nouns.

(15) a. A-ku-fik-anga muntu.] neg-17sm-arrive-pst.neg  person b. *A-ka-fik-anga] muntu. neg-1sm-arrive-pst.neg  person ‘Nobody came.’

Subjects of non-verbal predicates can appear in post-predicate position, as well. However, we will show that, with the exception of subjects of the existen-tial predicate -khona,these subjects are always dislocated and never predicate-internal. We will bring this in relation to a morphological property that distin-guishes non-verbal predicates from verbs. Specificallly, two morphemes, the negative marker-nga- and the persistive marker -sa-, are realized as -nge- and -se- respectively.

4 PredP and the non-verbal predication typesSince Bowers (1993) it has been assumed that there is a projection above the VP, which he called PredP, but which is now standardly called vP. Bowers also argued

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 9: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   431

that a similar projection was needed for non-verbal predication. This PredP7 is always present when an PP, AdjP or DP is used predicatively, including small clauses. Below is the structure for They consider John crazy, where there is a PredP, Johncrazy. This structure is taken directly from Bowers (1993):

(16) [IP Theyconsider [PredP John [Pred′ [Pred e ] [AP crazy]]]]

Bowers himself didn’t argue for a difference between the PredP dominating a VP and the one dominating an AdjP, PP or CP, but Baker (2003)argued that the PredP dominating a VP was different from the others, which we now call vP.

We will argue that in Zulu there is a difference between vP’s and PredP’s. This difference is reflected in the morphology. The negative marker -nga- and the per-sistive marker-sa-are realized as -nge-and-se-,8respectively, when they have a PredP as complement, as (17) shows.

(17) a. i-zingane ezi-nge-zincane det-10children  10rel-neg-10small ‘children who aren’t small’ b. i-zingane ezi-nga-cul-i det-10children  10rel-neg-sing-neg ‘children who don’t sing’ c. i-zingane ezi-se-zincane det-10children  10rel-prst-10small ‘children who are still small’ d. i-zingane ezi-sa-cul-a det-10children  10rel-prst-sing-fs ‘children who still sing’

There is also a syntactic difference between PredP’s and vP’s. With only one exception, specifiers of PredP, that is, subjects of non-verbal predication, always

7 The literature vacillates between the labels PredP and PrP. We have used the former uniformly throughout the paper, including when discussing analyses that use the former label.8 These forms also differ tonally. The verbal forms have a high tone, while the non-verbal ones have a falling tone. Falling tones have a very limited distribution in Zulu. The presence of a falling tone with non-verbal predicates is consistent with the theory, found in some traditional grammars such as Van Eeden (1956: 427), that these forms are diachronically -nga-and-sa- fol-lowed by the old copula li.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 10: Subject raising in Zulu and the nature of PredP

432   Leston Chandler Buell and Merijn de Dreu

move to a position relatively high in the tree, whereas specifiers of vP (that is subjects of verbal predicates) may remain in spec-vP, where they may receive an indefinite or a focused interpretation. As we will show, this difference can be traced to a more fundamental difference between PredP and vP. While a verb undergoes head-movement to v°, the analogous head of an AdjP, DP, or PP does not undergo head-movement to Pred°. We speculate that this may be due to some universal constraint on PredP’s that forbids head-movement to their heads. We know of no language where in nominal predication a determiner moves, leaving its NP complement stranded. Similarly, we know of no language in which an adposition may move, leaving its DP complement stranded in locative predica-tion. In adjectival predication it is less clear whether there is a universal or near-universal ban on head-movement, but we will assume that head-movement of adjectives to Pred° is also banned in Zulu. We will discuss DP’s, PP’s, and AdjP’s in the following sections, but we will start with verbal predication.

In verbal predication the verb moves to the v° position. The vP is dominated by a projection where the final suffix is located, the verbal suffix that varies with mood, aspect, tense, and polarity. Let’s call this projection the FSP. The complex head that consists of the v°, and the verb moves to FS° (Buell 2005; see also Muri-ungi 2009, Ch. 4 for a similar approach).

(18) 

As was diagrammed in (4), verbs can have a great deal of pre-stem morphol-ogy. This pre-stem morphology resides in projections dominating the FSP. This paper is not concerned with the question as to what the exact structure of the extended VP looks like. Our trees will thus have only one projection above FSP, labelled IP, and we will assume for the sake of simplicity that all pre-stem mor-phology resides there, although the actual structure is more complex. What is important for our discussion is that the verb remains in FSP, while the subject may either stay in spec-vP, or move to spec-IP, in the latter case triggering full noun class agreement on the subject marker:

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 11: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   433

(19) 

(20) 

It is the pre-verbal construction in (20) which serves as the basis for the post-verbal right-dislocation construction. The dislocated subject can be assumed to be base-generated in a high post-verbal position, while spec-IP is filled with a pro, coreferrent to the dislocated subject. An alternative analysis is that the subject itself moves to the higher position. Which of those two analyses is correct has no consequence for this paper.9 Both possibilities are captured in the structure in (21).

9 The tree in (21) does not follow Kayne’s (1994) widely accepted Linear Correspondence Axiom (“antisymmetry”), under which specifiers to the right of the head are forbidden. Were we to con-form to this antisymmetry, the dislocated subject would reside in a specifier of the TopP to the left of the IP, and then the IP would move to a specifier position above spec-TopP. For this paper there are no empirical differences between a consistently antisymmetric analysis and the one presented here.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 12: Subject raising in Zulu and the nature of PredP

434   Leston Chandler Buell and Merijn de Dreu

(21) 

In the following sections we will show that the structure equivalent to (19) is impossible in all but one type of non-verbal predication.

4.1 Adjectival and locative predication

We will start with the most straightforward cases; adjectival predication and loca-tive predication. Other types of non-verbal predication have additional properties that complicate the issue at hand. We will show that all post-predicate subjects in adjectival and locative predication are dislocated and never predicate-internal.

In the traditional Nguni literature (such as Doke 1973; Poulos and Msimang 1998; Van Eeden 1956), two subtypes of what here are termed “adjectives” have been treated separately. But for our purposes here, there is no relevant difference between agreeing and non-agreeing adjectives, traditionally (and somewhat con-fusingly) termed “adjectives” and “relatives”, respectively. These two subclasses of adjectives both bear relative morphology when used attributively (with the concomitant subject agreement morphology), but only the “agreeing” type has an additional adjectival agreement morpheme immediately preceding the stem. The two types of adjectives are illustrated attributively and predicatively in (22) and (23).10

(22) -ncane ‘small’ (agreeing adjective) a. Thina si-ba-ncane. 1pl.pron  1pl.sm-2-small ‘We are little/young.’ b. i-mifula e-mi-ncane det-4rivers  4rel-4-small ‘the little rivers’

10 For clarity’s sake, the adjectival prefix has been glossed separately in examples (22) through (24).

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 13: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   435

(23) -lusizi ‘sad’ (non-agreeing adjective) a. Thina si-lusizi. 1pl.pron  1pl.sm-sad ‘We are sad.’ b. i-mifula e-lusizi det-4rivers  4rel-sad ‘the sad rivers’

Note that an agreeing adjective bears two agreement prefixes. The leftmost one is the familiar subject marker, while the one to its right is an adjectival agreement prefix. In affirmative third person forms, the subject marker is usually absent, as in (24a), but the underlying presence of a subject marker can be dem-onstrated, for example, by exhibiting a negative counterpart, as in (24b).

(24) a. I-zindlovu zin-dala. det-10elephants  10-old ‘The elephants are old.’ b. I-zindlovu a-zi-zin-dala. det-10elephants  neg-10sm-10-old ‘The elephants aren’t old.’

In locative predication, just as in verbal and adjectival predication, there is no overt morpheme intervening between the TAM prefixes and the “head” of the predicate, in this case, the locative adverb or prepositional phrase. Locative expressions derived from noun phrases are often vowel-initial, and in this case an epenthetic s separates these two parts of the predicate word. An example of this predication type is given in (25):

(25) A-bafana ba-s-e-sikol-eni. det-2boys  2sm-epen-loc-7school-loc ‘The boys are at school.’

This s is taken to be epenthetic, rather than to be a locative predicate particle, for example, analogous to the copula in nominal predication, because the same seg-ment also appears after na ‘and’ in conjunctions, as in (26).

(26) lapha  na-s-e-sikol-eni here with-epen-loc-7school-loc ‘here and at school’

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 14: Subject raising in Zulu and the nature of PredP

436   Leston Chandler Buell and Merijn de Dreu

De Dreu (2008) and Mathonsi (2001) have analysed these locative constructions as PP’s with (s)e as their head.

While TAM morphology is attached directly to the locative phrase in affirma-tive forms, this is not possible in negative forms. Instead, a form with kho precedes the locative phrase, as in (27).

(27) A-bafana a-ba-kho e-sikol-eni. det-2boys  neg-2sm-loc.pred  loc-7school-loc ‘The boys aren’t at school.’

Post-predicate subjects in adjectival and locative predication are always dis-located. It can be demonstrated that they have the same properties as the dislo-cated subjects of verbal predicates. Such subjects cannot be focused, as shown in (28) and (29), and they cannot be an augmentless noun, as in (30):

(28) a. *Z-a-zincane] ziphi i-zinhlobo ze-zilwane? 10sm-pst-10small  10which  det-10kinds  10of:det-10animals b. Yi-ziphi i-zinhlobo ze-zilwane ez-a-zincane? cop-10which  art-10kinds  10of:det-10animals  10rel-pst-10small ‘What kinds kinds of animals were small?’

(29) *U-se-dolobhe-ni] u-Sipho kuphela. 1sm-loc-5town-loc  det-1a.Sipho  only intended: ‘Only Sipho’s in town.’

(30) a. *A-zi-ncane] zindlovu. neg-10sm-small  10elephants b. *A-ku-ncane zindlovu.] neg-10sm-small  10elephants intended: ‘No elephants at all are little.’

The conclusion that these subjects cannot be predicate-internal can be explained if there is no head-movement of the adjective or preposition to Pred°. The Head Movement Constraint (Travis 1984) prevents the adjective or preposi-tion from moving to FS°, because there is an intervening Pred° head. This means that if the subject remained in spec-PredP, it would intervene between the inflec-tion and the adjective or preposition. Since the inflection can only attach to the predicate, the subject must move to a higher position and cannot remain in spec-PredP. This inflection does not need to be overt; agreeing adjectives have a null morpheme in the 3rd person indicative present tense, as can be seen in (32).

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 15: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   437

(31) a. U-Sipho u-s-e-sikol-eni. det-1a.Sipho  1sm-epen-loc-school-loc ‘Sipho is at school.’ b. U-Sipho muhle. det-1a.Sipho  1beautiful ‘Sipho is beautiful.’

(32) 

This construction with a pre-predicate subject is the basis for the construc-tion with the post-predicate dislocated subject:

(33) 

Some agreeing adjectives have a different tone when negated. This tonal difference is reminiscent of the tonal component of the negative final suffixes and may be seen as evidence that adjectives do rise to FS°. However, these tonal differences may be alternatively explained in terms of stem selection (Buell 2006: §4.4.1.1).

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 16: Subject raising in Zulu and the nature of PredP

438   Leston Chandler Buell and Merijn de Dreu

4.2 Nominal predication

Nominal predication in Zulu makes use of a copular particle, which has three main segmental forms: ng(a)-, y(i)-, and w(u)-. In additional to these segmental forms, the copular morpheme can also often be expressed with depression11 (breathy voice) on the initial syllable of the nominal.12 Because depression is not indicated in the orthography, only segmental forms of the copula will be used here, even though the segmentless forms, when possible, are preferred in normal speech.

(34) U-Sipho (u)-ng-umfundisi. det-1a.Sipho  1sm-cop-1teacher ‘Sipho is a teacher.’

(35) Ngi-ng-umfundisi. 1sg.sm-cop-1.teacher ‘I am a teacher.’

Just as in adjectival predication, the subject marker in nominal predication is usually absent in principal submood when the subject is third person. However, the underlying presence of a subject marker is supported by the fact that it re-appears in participial submood.

There are significant dialectal differences in the negative forms of the para-digm. In the dialect described in Poulos and Msimang (1998), the TAM prefixes, including the prefix si- unique to these forms, are attached to an augmentless nominal, as in (36). Such forms will henceforth be termed “synthetic”. In contrast, our informant (a speaker of Durban Zulu) occasionally accepts a similar form without the si- morpheme as in (37a), but usually only accepts forms in which the TAM prefixes are attached to a pronoun which agrees with the following predicate nominal, as in (37b). These forms will be called ‘analytic’.

(36) Ka-ku-si-muntu lona,  yi-silwane. (synthetic) neg-17sm-si-1person  1this cop-7animal       ‘This isn’t a person; it’s an animal.’ (Poulos and Msimang, 1998, p. 362, adapted)

11 The segmental forms are also all depressed. The standard orthography does not distinguish between the modal and depressed variants of nasals, liquids, glides, or vowels.12 Poulos and Msimang (1998) spell this with hh, the digraph in the standard orthography for a voiced breathy glottal fricative.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 17: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   439

(37) a. À-kú-mùntù lónà. (synthetic) neg-17sm-1person  1this b. À-kú-yèná ú-múntù lónà.      (analytic) neg-17sm-1pron det-1person  1this ‘That’s not a person.’

As in verbal predication, post-predicate subjects of nominal predication are compatible with both a subject marker agreeing with the noun class of the subject or one bearing default class 17 agreement. One would expect these two choices to correspond with a dislocated subject and an internal subject, respectively. How-ever, the evidence shows that they are both dislocated. First, as shown in (38), both options are also available when the subject is in canonical preverbal posi-tion, while the class 17 option is not available for any other type of predication, including verbal predication.

(38) a. U-Sipho w-a-ye-ngu-mngane wethu. det-1a.Sipho  1sm-pst-1sm-cop-1friend  1our ‘Sipho was our friend.’ b. U-Sipho kw-a-ku-ngu-mngane wethu. det-1a.Sipho  17sm-pst-17sm-cop-1friend  1our

Buell (2011) and de Dreu and Buell (2013) argue that noun class 17 agreement of nominal predication agrees with a pro with noun class 17 features rather than with an expletive. They liken it to similar constructions in Dutch and in French, because the neuter Dutch pronoun het and French pronoun ce can only be used to refer to people in nominal predication. This is shown by contrasting nominal and adjectival predication in Dutch in (39).

(39) a. Die Peter toch! Hij/Het  is een  schat. that  Peter  indeed  it is  a treasure ‘That Peter! He’s a darling.’ (Buell 2012) b. Die Peter toch! Hij/*Het  is schattig. that  Peter  indeed  he/it is  precious ‘That Peter! He’s precious.’ (Buell 2012)

When the subject of nominal predication occurs in postverbal position, diag-nostics indicate that it is dislocated, regardless of the agreement pattern. As

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 18: Subject raising in Zulu and the nature of PredP

440   Leston Chandler Buell and Merijn de Dreu

observed in (40) and (41), regardless of the agreement option used, the subject cannot be modified with kuphela ‘only’ and cannot be a augmentless noun.

(40) a. *A-ku-muntu lona kuphela. neg-17sm-1person  1this  only intended: ‘Not only this one is a person.’ b. *W-a-ye-ngu-mngane wami  u-Sipho kuphela. 1sm-past-1sm-cop-1friend  1my det-1a.Sipho  only intended: ‘Only Sipho was my friend.’

(41) a. *A-ku-si-lona i-sela ntombazane. neg-17sm-si-5pron  det-5thief  9girl intended: ‘No girl is a thief’ b. *A-yi-si-lona i-sela ntombazane. neg-9sm-si-5pron  det-5thief  9girl intended: ‘No girl is a thief’

The post-predicate subject of nominal predication is thus always dislocated, even if the predicate has default noun class 17 agreement.

Our analysis for nominal predication is similar to our analysis of adjectival and locative predication. The predicate DP remains in its base position, while the subject must move out of the spec-PredP position because the inflection must attach to the predicate. What is different about nominal predication is that it makes use of a copular particle. Following Zeller (2010, 2012) we will assume that this particle is the realization of the Pred° head. The structure of nominal predica-tion with a pre-predicate subject is depicted in (42).

(42) 

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 19: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   441

4.3 Predication with na-

Many Bantu languages have a single morpheme which is used to mean ‘with’ ( comitative), which is used for coordination, and which serves as the basis for possessive predication. In Zulu, this morpheme is na, which is illustrated in each of these uses in (43a), (43b), and (43c).

(43) a. Ngi-cul-e na-lo mfana. 1sg.sm-sing-prf.cj  with-1this  1boy ‘I sang with this boy.’ b. Mina na-lo mfana  a-si-cul-anga. 1sg.pron  with-1this  1boy neg-1pl.sm-sing-neg ‘This boy and I didn’t sing.’ c. U-mfana u-na-marandi a-yikhulu. det-1boy  1sm-with:det-6rands  6rel-hundred ‘The boy has a hundred rand.’

This section is concerned with predicative uses of na, as in (43c). However, this type of construction actually has two distinct uses. It can be used for what is here straightforwardly called “possessive predication”, but with an expletive class 17 subject marker as in (44), it has an existential interpretation. We will discuss these two uses separately, starting with possession.

(44) Ku-na-marandi a-yikhulu. 17sm-with:det-6rands  6rel-hundred ‘There are a hundred rand.’

As with the three other types of predication just discussed, the subject of predication cannot be left in a position below the inflectional domain. This would result either in the possessor intervening between inflection and na- or between na- and the possessum. The examples in (44) show that both configurations are ungrammatical:

(45) a. *Ku-no-Sipho i-ncwadi. 17sm-with:det-1a.Sipho  det-9book b. *Ku-(u)Sipho ne-ncwadi. 17sm-det-1a.Sipho  with:det-9book intended: ‘Sipho has a book.’

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 20: Subject raising in Zulu and the nature of PredP

442   Leston Chandler Buell and Merijn de Dreu

Also as in the predication types already discussed, the subject may follow the entire predicate, but in that case it is clearly dislocated. This is shown in (46). The fact that lo mfana cannot be modified by kuphela, and that it cannot be an augmentless noun shows that it is dislocated.

(46) a. U-ne-ncwadi lo mfana. 1sm-with:det-9book  1this  1boy ‘He has a book, that boy.’ b. *U-ne-ncwadi lo mfana  kuphela. 1sm-with:det-9book  1this  1boy only intended: ‘Only this boy has a book.’ c. *A-yi-na-ncwadi13 ntombazane. neg-9sm-with-9book  9girl intended: ‘No girl has a book.’

As shown in (47), the structure of this construction with na is the same as the structure of locative predication. The PP remains in situ and the subject moves to a higher position because the inflection attaches to the predicate and not to the subject.

(47) 

Now we turn to existential predication with na-. The difference with posses-sive predication with na is that the subject agreement is always noun class 17 ku- and the only argument is the complement of na.An example is given in (48):

13 Complements of na in negative possessive and existential predication do not have the augment.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 21: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   443

(48) Ku-ne-mali e-ningi e-lahlekile. 17sm-with:det-9money  9rel-9much  9rel-lost:dj ‘There’s a lot of money that has been lost.’

This argument is always internal. (49a) shows that it can be a wh phrase, (49b) shows it can be focused, because it can be modified by kuphela,and (49c) shows it can be an augmentless noun:

(49) a. Ku-na-ziphi i-zinhlobo ze-zihlahla lapha? 17sm-with-10which  det-10types  17of:det-10plants  here ‘What kinds of trees are there here?’ b. Ku-ne-zihlahla kuphela  lapha. 17sm-with:det-10plants  only here ‘There are only trees here.’ c. A-ku-na-mali eningi e-lahlekile. neg-17sm-with-9money  9much  rel:9sm-lost:dj ‘There is not a lot of money that is lost’

This probably doesn’t constitute a counterargument to our idea that subjects of non-verbal predication are always dislocated because it is questionable if the argument in question is a subject at all. Rather, it is probably a complement of na, while the subject is an expletive. It is comparable to German esgibt,where the subject is an expletive es,and the only argument is a direct object, in the accusa-tive, as is shown by the case of the indefinite article (50):

(50) Gibt es  einen Weihnachtsmann? gives  it a.m.acc  Santa.Clause ‘Is there a Santa Clause?’

The tree structure of this is in most ways similar to the tree structure of pos-sessive predication. The only difference is that there is no subject. Thus there is nothing to raise to a higher position.

(51) 

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 22: Subject raising in Zulu and the nature of PredP

444   Leston Chandler Buell and Merijn de Dreu

4.4 Existential predication with -kho(na)

Another way of expressing existential predication is with -kho(na). The form is -khona in affirmative contexts, and simply kho in negative contexts. It is exempli-fied in (52):

(52) a. Ku-khona i-mali e-ningi e-lahlek-ile. 17sm-loc.pred  det-9money  9rel-9much  9rel-get.lost-prf.dj ‘There is a lot of money lost.’ b. A-ku-kho mali e-ningi e-lahlekile. neg-17sm-loc.pred  9money  9rel-9much  9rel-get.lost-prf.dj ‘There isn’t a lot of money lost.’

-Kho(na) can have either subject agreement that agrees in noun class with the subject or default noun class 17 agreement. Post-predicate subjects that trig-ger full agreement are dislocated, but if the subject agreement is the default 17 they are internal. This can be seen in (53), in which the subject is modified by kuphela, in (54), in which the subject is questioned, and in (55), in which the subject is augmentless. In all three cases the result is grammatical only if the verb bears default noun class 17 agreement rather than full noun class agreement:

(53) a. Ku-khona i-jusi kuphela.] 17sm-loc.pred  det-5juice  only b. *Li-khona] i-jusi kuphela. 5sm-loc.pred det-5juice  only ‘There’s only juice.’

(54) a. Ku-khona ziphi i-zinhlobo ze-mithi] lapha? 17sm-loc.pred  10which det-10kinds 10of:det-4medicines here b. *Zi-khona] ziphi i-zinhlobo ze-mithi lapha? 10sm-loc.pred  10which  det-10kinds  10of:det-4medicines  here ‘What kinds of medicines are there here?’

(55) a. A-ku-kho mali eningi.] neg-17sm-loc.pred  9money  9much ‘There isn’t a lot of money.’ b. *A-yi-kho] mali eningi. neg-9sm-loc.pred  9money  9much ‘There isn’t a lot of money.’

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 23: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   445

Both of our diagnostics thus indicate that a lexical subject argument with this predication type implies that it is left-dislocated. This would thus appear to be a genuine counterexample to our proposal that post-predicate subjects are always dislocated.

This anomaly can be explained if we assume that -khona behaves just like a verb in that it can undergo head-movement to the Pred° head and from that posi-tion further to the head of FSP. Under these assumptions, a subject can remain in spec-PredP.

(56) 

Additional evidence that -kho(na) undergoes head-movement comes from relative clauses in closely related Xhosa. Verbs in relative clauses in Zulu and Xhosa sometimes have a suffix of the form -yo that follows the final suffix, shown in a Zulu relative clause in (57a). An analysis of this suffix has been sketched out in Buell (2006) for Zulu, though the Xhosa particle differs from its Zulu counter-part in that it is employed in a few additional environments. There are two pos-sible positions this suffix might reside in; one is the specifier of a projection below FSP, and one is as a head of a projection above FSP, which we can call RelP, while the whole verb complex including the final suffix adjoins to Rel°. If we follow the latter analysis we can use that as evidence that there is head movement in verbal predication. Only if there is there has been movement to FS° can there be head movement to Rel°. A simplified tree structure is given (57b).

(57) a. u-mfana o-cul-a-yo det-1boy  1rel-sing-fs-yo ‘the boy who sings’

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 24: Subject raising in Zulu and the nature of PredP

446   Leston Chandler Buell and Merijn de Dreu

b.

In Xhosa -yo can attach to -kho as well, as shown in (58a). This strongly sug-gests that there is also head movement of this element in Xhosa, as depicted in (58b).

(58) a. u-mntu o-kho-yo det-1person  1rel-present-yo ‘a person who is present’ (Mini et al. 2003: 92)

b.

-Khona thus behaves like a verb in that it must undergo head-movement, leading to the question as to why, given that no other non-verbal predication heads do so. We would like to suggest that it is, in fact, a verb, but a special one: one that is dominated by PredP rather than vP. All its non-verbal features, such as the special durative and persistive allomorphs -nge- and -se-, are dependent on the presence of PredP instead of vP, whereas -khona’s need to undergo head-movement may be dependent on it being of the category of verb.

Alternatively, Zeller (2010) argues that -khonais the realization of Pred°. This could also explain why it can undergo head-movement; if it is base-generated in Pred° the head-movement constraint does not apply since there is no head in

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 25: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   447

between Pred° and FS°. We remain agnostic about the exact analysis of -khona, but argue that it undergoes head-movement to FS° whatever the exact status of -khona is.

5 Verbs in non-verbal predication

We have called all cases where no lexical verb appears “non-verbal predication”. However, in inchoative non-verbal predicates the verb -ba does in fact appear. In addition to the case of -ba, some scholars (Posthumus 2000a,b) have argued that a verb, referred to for historical and cross-Bantu reasons as -li, is always present in stative non-verbal predicationand that it is this verb that is responsible for the forms -nge-and -se- found in our “non-verbal” predication. That is, under that analysis these forms are the product of fusion or vowel coalescence, as in (59), in which the actual underlying form of the verb is assumed to be -i. Note that adja-cent vowels are illicit word-internally in Zulu.14

(59) /-ngà- + -ì-/→ [ -ngè-] /-sá- + -ì-/→ [ -sê-]

In this section we will first examine sentences with -baand show that their structure is essentially the same as those without -ba. Then we will show that an analysis with an underlying-li is not superior to our own analysis, which relies on allomorph selection. Finally, we will also show that even if there were an under-lying -li responsible for the -nga-/-nge- and -sa-/-se- alternations, its presence would make no difference for our larger point, which is that the form of -nga-/-nge- and -sa-/-se- is ultimately dependent on the presence of vP or PredP.

5.1 The auxiliary -ba

In addition to the stative morpheme -beshown in examples (9) through (11), in clauses with non-verbal predication sometimes a form of etymologically related

14 In vowel-initial nouns, such as u-anti ‘aunt’, a glottal stop is pronounced between the stem and any prefixes, such as an augment.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 26: Subject raising in Zulu and the nature of PredP

448   Leston Chandler Buell and Merijn de Dreu

-ba appears, which we will consider to be an auxiliary. In some contexts -ba is obligatory, such as in infinitives, imperatives, and subjunctives. In other contexts constructions with -baoccur, as well as constructions without -ba.In those con-texts the constructions with -ba have the interpretation of ‘become’ in nominal and adjectival predication. In other types of predication, sentences with -ba and without are both interpreted as stative, while speakers perceive a subtle differ-ence between them which is hard to characterize (Jochen Zeller, p.c.). (60a) is an example of -ba with nominal predication.

(60) a. Ngi-yo-ba ngu-dokotela ngelinye  ilanga. 1sg.sm-rem.fut-ba  cop-1a.doctor  on:5one 5day ‘I will become a doctor one day.’ b. Ngi-yo-be ngi-ngu-dokotela lapho  ngi-phuma 1sg.sm-rem.fut-be  1sg.sm-cop-1a.doctor  there 1sg.sm-leave phesheya. abroad ‘I will be a doctor when I return from overseas.’ (Poulos and Msimang 1998: 365, glosses our own)

In other tenses than the present tense, this construction is very similar to the normal stative copular construction in non-present tenses with -be, exemplified here by (60b). However, an additional morphosyntactic phenomenon is contin-gent on the choice between -be and -ba. Note that in (60b) -be followed by ngi- on the predicate nominal. Inchoative -ba is never followed by the subject agreement, as shown in (60a).

Whatever -ba is, either directly or indirectly it selects a PredP as its comple-ment: it is never followed by a verb,15 and in nominal predication the noun is preceded by the copular particle, which we argued to be the realization of PredP. Any final suffix follows -ba. So, -bais most likely either base-generated as a head between FSP and PredP, in which case it moves to FSP, or it is base-generated in a position right above FSP. Below are trees illustrating the two analyses, where we have called the projection headed by -ba InchP (for “Inchoative Phrase”).

15 We differentiate between -ba, which is never followed by a verb, and the etymologically related -be, which we discussed in Section 2 and which can be followed by a verb.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 27: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   449

(61) a.

b.

Internal subjects are not allowed in non-verbal predicates with -ba:

(62) a. *Ku-yo-ba ngu-mfundisi  u-Sipho kuphela. 17sm-rem.fut-ba  cop-teacher det-1aSipho  alone intended: ‘Only Sipho will be a teacher.’ b. *Ku-yo-ba muhle u-Sipho kuphela. 17sm-rem.fut-ba  beautiful  det-1aSipho  alone intended: ‘Only Sipho will be beautiful.’ c. *Ku-zo-ba e-Thekw-ini u-Sipho kuphela. 17sm-fut-ba  loc-Durban-loc  det-1aSipho  alone intended: ‘Only Sipho will be in Durban.’

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 28: Subject raising in Zulu and the nature of PredP

450   Leston Chandler Buell and Merijn de Dreu

This can be explained if we assume that -ba has the same need to attach to the predicate as the inflection does. There is some additional evidence for this. In older grammars (e.g., Van Eeden 1956), -ba is often written as a single word with the following predicate, and this is still the case when the predicate is phonologi-cally light, such as -kho is in (63). In this latter case, the writing convention re-flects the stress facts. Zulu has regular penultimate stress, and the fact that the stress falls on bi in ukungabikho in (63) shows that -kho has been incorporated into the word phonologically. These facts suggest that -ba is not phonologically independent:

(63) u-ku-nga-b-i-kho kwe-zindlu16

det-15inf-neg-ba-neg-kho  15of.det-10houses ‘the absence of houses’

That -ba forms one phonological word with the following predicate is also shown by the form of the locative marker (-s)e-. As discussed in Section 4.1, the s is epen-thetic. It doesn’t occur when the locative marker follows a word boundary, but it is used word-internally to prevent either vowel adjacency or vowel coalescence. When the locative marker follows -ba, it always has the epenthetic s, as shown in (64), suggesting that -ba- and the following locative marker are part of the same phonological word.

(64) Ngi-zo-ba se-Thekw-ini. *eThekwini 1sgs-fut-ba  loc-Durban-loc       ‘I will be in Durban.’

Thus, because -ba is phonologically dependent, the same mechanism that rules out internal subjects in stative predicates also rules out internal subjects in incho-ative predicates.

5.2 The presence of -li in non-verbal predication

In many other Bantu languages, clear reflexes of a copular verb -li appear in some environments. For example, in Swahili, -li shows up in affirmative tense relative clauses with non-verbal predicates, as in (65b):

16 Found at http://www.kznhousing.gov.za/Portals/0/docs/Speeches/2008/Speech%20-%20KwaMaphumulo.pdf.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 29: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   451

(65) a. Ni-na pesa. 1sg-with  10money ‘I have some money.’ b. Pesa ni-li-zo na-zo zi-ta-tosh-a. 10money  1sg-li-10pron with-10pron  10sm-fut-suffice-fs ‘The money that I have will be enough.’

In another Bantu language, Southern Sotho, the form -leis used in the present tense participial, which is used after leha ‘although’, amongst other uses:

(66) a. lēha kē-lē mōbē although  1sg-le  1bad ‘although I am bad’ (Doke & Mofokeng 1985: 303)

It has been argued (Posthumus 2000a,b) that this verb is also present in Zulu, but that it is never pronounced as a discrete element, its only effect being the raising of the vowel of -nga-and -sa-. Some evidence (not discussed by Posthu-mus) for this claim is that -sa-differs tonally from -se-. The verbal form -sá- is pronounced with a high tone, while the non-verbal -sê- has a falling tone. Falling tones have a very limited distribution in Zulu. In a few cases a falling tone arises as a result of a contraction of two syllables. For instance, in colloquial speech the z of the class 10 class prefix zì- can be elided when preceded by the augment í-, yielding a single falling vowel î. For example, the word ízìngánè ‘children’ can colloquially be pronounced îngánè. With regards to -sa-/-se-, we assume that the falling tone of -se- before non-verbal predication is likewise the result of a diachronic contraction with the historical -li, as proposed in several traditional grammar books such as Van Eeden (1956). However, this diachronous change does not mean that a synchronic underlying -li should necessarily be assumed.

Another important piece of evidence for the presence of a verb -li is that if both -nga-and -sa- are present, then only the latter one, -sa-, manifests the raised vowel, as shown in (67).

(67) u-buso obu-nga-se-buhle *obungesebuhle det-14face  14rel-neg-prs-14beautiful       ‘a face which is no longer beautiful’

In our analysis of allomorph selection we would have to resort to an ad hoc expla-nation, keeping everything in the analysis the same for -sa-/-se-, but stipulat-ing  that -nga- selects either for vP or for -sa-/-se-. Unfortunately, however, the

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 30: Subject raising in Zulu and the nature of PredP

452   Leston Chandler Buell and Merijn de Dreu

explanation relying on an underlying -li also necessarily suffers from certain ad hoc elements itself. If there were an unpronounced -li present in non-verbal pred-ication, we would expect all inflectional morphemes ending in an -a to be affected. But this is not the case. Both the noun class 6 subject marker -(w)a-and the noun class 1 negative subject marker -ka- remain unaffected, as seen in the following examples:

(68) a. U-Sipho a-ka-mubi. *akemubi det-Sipho  neg-1sm-1ugly     ‘Sipho isn’t ugly.’ b. A-madoda  a-se-Thekwini. *eseThekwini det-6men 6sm-loc-Durban       ‘The men are in Durban.’

Thus, the effect of putative -lion -nga-and -sa- is not a purely phonological matter. Some other mechanism or stipulation is needed to ensure that -nga-and -sa- are the only morphemes affected.

An additional problem with assuming an underlying -li is that if the predicate is the negative -kho, the noun class 6 subject marker -(w)a-and noun class 1 neg-ative subject agreement -ka- are affected, as seen in the following examples:

(69) a. U-Sipho a-ke-kho. det-Sipho  neg-1sm-kho ‘Sipho isn’t there.’ b. A-madoda  a-we-kho e-Thekwini. det-men neg-6sm-kho  loc-Durban ‘The men aren’t in Durban.’

This would be puzzling under the -lianalysis. We would expect -li to always have the same effect and to always affect the same morphemes regardless of what its complement is. Conversely, if we say that the alternations are the result of allo-morphy, we can simply say that the forms -(w)e- and -ke- select -kho.

5.3 The consequences of -li for our analysis

As we argued, an analysis with underlying -li is not superior to an analysis with allomorph selection. However, even if we were to assume an underlying -li in the case of non-verbal predication, this would not affect our larger point, which is that the form of -nga-/-nge- and -sa-/-se- is dependent on whether PredP or vP is

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 31: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   453

present. Since putative li- would only occur with non-verbal predication,17 it would only be present if there were a PredP rather than a vP. Like -ba-, -li would be neither a lexical verb and nor of category V, but would rather be an auxiliary, one not dominated by vP. We are not sure what the exact category of -li would be, but as is shown in the tree below, where it is given the label Cop°, its projection would dominate PredP.

(70) a. (Ku-fanele u-hamb-e u-bon-e umhlaba) 17.sm-should  2sg.sm-leave-sbjv  2sg.sm-see-sbjv  3world u-se-musha. 2sg.sm-still-1young ‘(You should go see the world) while you are still young.’

b.

So, to reiterate, sentences with -ba can be analysed the same way as sen-tences without it,because -ba is phonologically dependent, just like the inflec-tion. There is no clear reason to prefer an analysis with an underlying -li to one with allomorph selection, and even if there were an underlying -li its presence would depend on whether it selects a vP or a PredP. Therefore, the choice between these two analyses makes no difference to our larger point that the form of -nga-/-nge-and -sa-/-se-is ultimately dependent on the presence of a vP or PredP.

17 A very small number of verbs also have the alternative forms -nge- and -se-. However, the situation is different for these verbs because in any morpheme with an /a/ preceding the stem of one of these verbs the /a/ is raised.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 32: Subject raising in Zulu and the nature of PredP

454   Leston Chandler Buell and Merijn de Dreu

6 Predicate type and the nature of PredPWe will discuss the PredP projection in more detail presently, but it will be useful to first recapitulate the proposed analysis and the facts it captures. (We exclude -khona here, whose anomalous properties were discussed in Section 4.4.) There are two main facts which need to be accounted for.

First, there is a morphological alternation with respect to the persisitive -sa-/-se- and negative -nga-/-nge- prefixes. When combined with verbs one form appears, and when combined with non-verbal predicates another form is re-quired. This is accounted for by assuming different categories capping off the thematic domain. With a verb, that category is vP; with other predicate types, it is PredP. The morphology is sensitive to the category of the predicate.

Second, inversions with predicate-internal subjects are only possible with verbs. This is attributed to the divergent morphological properties of v° and Pred°, and to the verb’s need to move at least as high as to FS°. While V° can adjoin to v°, and indeed must do so to move further to adjoin to FS°, neither Adj°, P°, nor D° can adjoin to Pred° in a similar fashion. The two situations are schematized in (71).

(71) a. [IP inflection [FSP verbi [vP subject [v′ ti [VP ti ]]]] b. *[IP inflection [PredP subject [AdjP adjective ]]]

It is the verb’s adjunction to FS° that makes the VS word order in (71a) possible: the verb moves above the subject, which remains in spec-vP. As for (71b), the reason for its ungrammaticality is morphological: inflection must attach to a predicate, be it a PredP or a vP. It cannot attach to an intervening DP.

We follow the majority of work on non-verbal predication since Bowers (1993) in assuming there is a functional category dominating AdjP’s, DP’s, and PP’s in whose specifier the subject is base-generated. Further, we follow Baker (2003) in assuming a distinction between PredP and vP. There is strong morphological evidence for PredP in Zulu. As already discussed, in nominal predication there is a copular particle which has different forms depending on noun class and cate-gory of the predicate. We assume, following Zeller (2010), that this particle is an overt manifestation of Pred°.

6.1 On the complementarity of PredP and vP

The idea that a non-verbal predicate’s inability to appear in Zulu inversions is attributable to the projection of PredP was first proposed by Zeller (2010) and

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 33: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   455

Zeller (2012) (where he uses the label “PrP”). The details of these two papers vary to some extent. However, both papers differ from ours in that Zeller argues that PredP projects above a wider range of predicate types than we do. Here we summarize the most essential syntactic aspects of these two papers as a unified analysis.

With respect to non-verbal predication, Zeller’s account is essentially the same as our own. Pred° selects an AdjP, PP, or NP. The subject of predication is base-generated in spec-PredP. Just as in our own analysis, the copular particle is assumed to be an overt Pred° head when the selected category is an NP. The argument in spec-PredP raises to spec-IP.

In addition to non-verbal predicates, Zeller argues that PredP is projected in two classes of constructions in which the verb precedes the external argument (e.g., the agent). The first of these is locative and instrumental inversion, illus-trated in (72b) and (73b).

(72) a. unmarked word order A-bantu aba-dala ba-hlal-a ku-lezi zindlu. det-2people  10rel-old  2sm-stay-fs  in-10these  10houses ‘People live in these houses.’ b. semantic locative inversion18

Lezi zindlu zi-hlal-a a-bantu a-badala. 10these  10houses  10sm-stay-fv  det-2people  2rel-2old ‘These houses are occupied by old people.’ (literally ‘These houses live

old people.’) (Buell 2007: 111)

(73) a. unmarked word order U-mfundi u-bhal-a nge-peni. det-1student  5sm-write-fs  with:det-5pen ‘The student is writing with a pen.’ b. instrumental inversion I-peni li-bhal-a u-mfundi. det-5pen  5sm-write-fv  det-1student19

‘The student is using a pen to write.’ (Zeller, 2012)

18 The term “semantic locative inversion” (in which locative expression appears in canonical form) is due to Buell (2007), who contrasts it with “formal locative inversion” (in which the loca-tive expression appears in locative form).19 We have adapted the glosses of Zeller’s examples to be consistent with our own glossing.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 34: Subject raising in Zulu and the nature of PredP

456   Leston Chandler Buell and Merijn de Dreu

In these constructions it is not the logical subject that triggers subject agreement, but rather a DP that semantically functions as an adjunct and that in a non- inverted construction would be expressed as an oblique (i.e., ‘in these houses’, ‘with a pen’). In locative and instrumental inversion, PredP selects a vP. The logical subject (e.g., the agent) remains in spec-vP. The locative or instrumental argument is generated in spec-PredP. In Zeller (2012) the relation between the DP occupying spec-PredP and the argument structure within vP is kept vague, while Zeller (2010) assumes that the DP binds a gap inside vP. Again, the argument in spec-PredP raises to spec-IP.20 The analysis as applied to (73b) is sketched in (74).

(74) 

The second type of construction beyond non-verbal predication that Zeller assumes to involve a PredP is the expletive subject clauses with VS word order and expletive class 17 subject agreement on the verb first encountered in Section 2, such as the one in (75) (repeated from (12a)).

(75) Ku-fik-e i-zingane. 17sm-arrive-prf.cj  det-10children ‘The children have arrived.’

In this case, the specifier of PredP is an expletive pro with noun class 17 features. This pro subsequently moves to spec-IP, triggering subject agreement on the sub-ject marker, as sketched in (76).

20 This analysis is similar to that of Buell (2005, Ch. 7), who, like Zeller, argues that the locative argument in this type of locative inversion (since Buell [2007] often called “semantic locative inversion”) is introduced above the external argument. The two analyses differ in that Buell argues that the projection selecting vP is an ApplP headed by the overt applicative suffix, while Zeller argues that it is PredP, headed by a phonologically empty morpheme.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 35: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   457

(76) 

A syntactic point these three cases have in common is that the specifier of PredP must always move to spec-IP. Only arguments in positions below PredP can be left in situ. This accounts for the fact that there are no predicate-internal subjects with non-verbal predication, because the subject of predication is never in a posi-tion below PredP where it could be stranded. Furthermore, it seems to be implied that leaving an argument in spec-vP is only possible when PredP is projected.

Zeller deviates from Bowers (1993) in assuming that PredP does not project in every case of verbal predication. Bowers assumes that there is a PredP in all cases of verbal predication as well as with non-verbal, and in fact, Bowers’s conception of PredP has evolved into the now familiar vP in the syntactic literature.Zeller (2012) deviates from this development by arguing for the co-occurrence of PredP and vP. This amounts to assuming a position distinct from Bowers’s original PredP. As far as we know, everyone who has assumed that there is a PredP since Bowers either assumes it to be used in verbal predication and sees it as another term for vP (Hazout 2004), or assumes it to be in complementary distribution with vP (Baker 2003).

The most salient difference between Zeller’s analysis and our own lies in the range of constructions in which PredP is projected, and we can therefore refer to Zeller’s analysis as involving “Permissive Projection” of PredP, while ours involves “Restricted Projection” of same. In the following section we discuss the nature of internal subjects under the two analyses, but first it should be pointed out that the two analyses cannot account for the same range of phenomena. We claim that the presence of PredP is what triggers the durative and negative allomorphs -se-and -nge-. However, in locative inversion, the durative and negative prefixes appear as -sa- and -nga-, just as in the cases where neither we nor Zeller assume the presence of a PredP. This is shown with the durative prefix in (77).21

21 We expect that instrumental inversion will also employ the -sa- and -nga- allomorphs. Unfortu-nately, we cannot test this, as our informant does not accept the instrumental inversion construction.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 36: Subject raising in Zulu and the nature of PredP

458   Leston Chandler Buell and Merijn de Dreu

(77) Lezi zindlu zi-sa-hlal-a abantu a-badala. *zisehlala these10 10houses  sm10-prs-stay-fs  2people  2rel-2old     ‘Old people still live in these houses.’

The idea that PredP is what triggers the -se- and -nge- allomorphs is thus in compatible with the idea that PredP is what licenses locative and instrumental inversion.

6.2 The nature of internal subjects

As explained above, according to Zeller, PredP is projected even with a verbal predicate if there is an internal subject. In that case he assumes that there is an expletive pro in the specifier of PredP. It is this expletive pro that triggers subject agreement upon moving to preverbal subject position. According to him, this explains why in Zulu an internal subject always has the presentational focus he claims it to have. The semantics of Zulu constructions with internal subjects is claimed to be similar to that of existential predication in other languages. He follows Hazout (2004) in assuming that existential predication is the result of having a pro expletive in the specifier of PredP. This operation turns the comple-ment of PredP into a syntactic predicate. This syntactic predicate is semantically interpreted as existential predication, with the complement of PredP as its argu-ment. This complement also has presentational focus. Zeller argues that some-thing similar happens with internal subjects. Here a PredP is projected whose expletive pro specifier forces the subject to be interpreted as having presenta-tional focus. In this section we will argue that this analysis fails to account for interpretational similarities between internal subjects and internal objects, and that actually the internal subject does not always have presentational focus in this configuration.

Before we examine the semantic properties of internal subjects, we first wish to point out something that can been viewed as a weakness in Zeller’s argumenta-tion. As just explained, he follows Hazout (2004) in assuming that if a DP is the complement of PredP, and if the specifier of PredP is an expletive, then its com-plement DP is syntactically a predicate and the resulting interpretation is that of an existential predicate with the DP as its argument. Zeller extends this analysis by assuming that something similar happens with internal subjects, where there is an expletive in the specifier of PredP, and the subject has presentational focus. However, in the case of internal subjects, the complement of PredP is not the DP but a vP or VP. Thus, a property first attributed to the complement of Pred° is now being attributed to the specifier of the complement. This is a big jump whose justification seems questionable.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 37: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   459

We now turn to the properties of the internal subjects themselves. Zeller claims that PredP is responsible for the focus interpretation of Zulu internal subjects. However, in addition to having internal subjects Zulu also has internal objects. These internal objects can be distinguished from external (i.e., displaced) objects by their lack of object agreement, which takes the form of an object marker on the verb. Objects can be shown to be vP-internal by application of the same tests that show that internal subjects are vP internal, shown in examples (78) through (80). In all of these examples, an object marker before the verb stem in-dicates that the object is displaced (as does the use of a disjoint verb form in (79) and (80)) (Van der Spuy 1998; Buell 2005). (78) shows that only an internal object can lack the augment, which results in an emphatic NPI (negative polarity item) interpretation. Using modification with kuphela ‘only’, (79) demonstrates that only an internal object can be focused. And finally, (80) shows that only an inter-nal object can be a wh constituent. In these aspects, the distribution of internal and external objects is identical to that of internal and external subjects. Internal and external objects thus have the same range of interpretations as their internal and external subject counterparts. Zeller, however, cannot explain this fact, be-cause for him it is the projection of PredP that makes the pragmatic interpretation of the internal subject possible.

(78) a. A-ngi-bon-i muntu.]      (Halpert 2012) neg-1sg.sm-see-neg  1person ‘I don’t see anyone.’ b. *A-ngi-m-bon-i] muntu. neg-1psm-1om-see-neg  1person ‘I don’t see anyone.’

(79) a. *Ngi-m-bon-e u-Sipho kuphela.]VP

1sg.sm-1om-see-prf.cj  det-1a.Sipho  only ‘I saw only Sipho.’ (Buell 2008) b. *Ngi-m-bon-ile]VP u-Sipho kuphela. 1sg.sm-1om-see-prf.dj  det-1a.Sipho  only intended: ‘I saw only Sipho.’ (Buell 2008)

(80) a. U-kheth-e liphi i-hembe?] 2sg.sm-choose-prf.cj 5  which  det-5shirt ‘Which shirt did you choose?’ b. *U-li-kheth-ile] liphi i-hembe? 2sg.sm-5om-choose-prf.cj  5which  det-5shirt

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 38: Subject raising in Zulu and the nature of PredP

460   Leston Chandler Buell and Merijn de Dreu

Zeller claims that his analysis explains why internal subjects have presenta-tional focus. However, on closer inspection, the assertion that all internal sub-jects have presentational focus turns out not to hold. For instance, when a subject or object has contrastive focus (as opposed to presentational focus), it must also be internal, as in the contrastively focused internal subject in (81).

(81) Ku-cul-e u-Sipho,]VP hhayi  u-Thandi. 17sm-sing-prf.cj  det-1a.Sipho  not det-1a.Thandi ‘SIPHO sang, not Thandi.’

It is also questionable whether focused arguments that are modified by kuphela ‘only’ or that are wh phrases have presentational focus. One piece of evidence to the contrary is clefts. If internal subjects always had presentational focus, and if this were a result of them being a part of an existential construction, then we would expect to find cleft sentences with existential predication in the higher clause. However, when a wh phrase or a phrase modified by kuphela is clefted, the standard cleft is used. Thus, the best paraphrase of the internal subject question in (82a) is (82b), in which the upper clause employs nominal predication. The ungrammaticality of (82c) and (82d) show that paraphrases with -khona and kuna- in the matrix clause are not possible.

(82) a. Ku-cul-e ziphi i-zingane? 17Sm-sing-prf.cj  10which  det-10children ‘Which children sang?’ b. Yi-ziphi i-zingane ezi-cul-ile? cop-10which  det-10children  10rel-sing-prf.dj ‘What kinds of animals were small?’ c. *Ku-khona ziphi i-zingane ezi-cul-ile? 17sm-loc.pred  10which  det-10children  10rel-sing-prf.dj ‘Which children are that sang?’ d. *Ku-na-ziphi i-zingane ezi-cul-ile? 17sm-with-10which  det-10children  10rel-sing-prf.dj ‘Which children are that sang?’

Taken together, these facts seriously damage the semantic support of Permissive Projection of PredP.

We argue, following Buell (2008) and Cheng and Downing (2007), that both internal subjects and internal objects have a range of interpretations that can be best summarized as not being topics or as not being “given”, and it is topicaliza-

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 39: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   461

tion that moves a subject or object out of the vP in normal sentences. It seems that subjects in SpecIP or higher have topic-like properties. The external argument remains in its in situ spec-vP position when its intended interpretation is incom-patible with topicality, as is necessarily the case when it is focused. No additional mechanism is needed to explain the focus. Zeller claims that PredP changes the perspective of a sentence by adding an argument to which the vP is predicated. As a response to this idea, we would say that this is precisely the function of the preverbal subject position in Zulu. That is, the preverbal subject position is what determines what the sentence is about. In those cases where the subject is exple-tive the sentence is thetic. That is, the sentence is not about anything. In this way, no additional mechanism is needed to account for the change of perspective char-acterizing locative and instrumental inversion.

Zeller does have an additional argument for assuming a PredP with internal subjects. He claims that he can explain the fact that internal subjects in expletive subject constructions, locative inversion, and instrumental inversion have a similar distribution: these constructions are incompatible both with non-verbal predicates and with object agreement. The examples in (83) show that locative inversion and internal subjects are not possible with non-verbal predication,22 whereas (84) shows that object agreement is ungrammatical when the verb is transitive.23

(83) a. *Lo mfula a-wu-m-khulu  (i-)zinhlanzi. 3this  3river  neg-3sm-3big (det-)10fish intended: ‘This river is where the fish aren’t big.’ (Zeller 2010) b. *A-ku-ncane zindlovu.] neg-10sm-small  10elephants intended: ‘No elephants at all are little.’

(84) a. *I-sipunu si-li-dl-a u-John. det-7spoon  7sm-5om-eat-fs  det-1a.John intended: ‘John is using the spoon to eat it (e.g. the soup).’ (Zeller 2012)

22 Zeller does not discuss the question of instrumental inversion with non-verbal predication, but it follows from his analysis that it is impossible. While we expect that to be true, we cannot test it because our informant does not accept instrumental inversion.23 Buell (2005, ch. 5) shows that the reflexive marker can appear in the expletive subject con-struction, in contrast to other object markers.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 40: Subject raising in Zulu and the nature of PredP

462   Leston Chandler Buell and Merijn de Dreu

b. *Lesi  sikole si-lu-fund-el-a a-bantwana. 7this 7school  7sm-11om-study-appl-fs  det-2children intended: ‘Children study it at this school.’ (Zeller 2010) c. *Ku(-ya)-yi-phek-a u-mama ekhish-ini. 17sm-(dj)-9om-cook-fs  det-1a.mother  loc-5kitchen-loc intended: ‘Mother is cooking it in the kitchen.’ (Zeller 2010, modified)

There are, however, cases where internal subjects are allowed but cannot undergo locative inversion. One of these is with -khona. While an internal subject is pos-sible with this predicate, locative inversion is not:

(85) a. Ku-khona a-bafundi ku-lesi sikole. 17sm-loc.pred  det-2students  prep-7this  7school ‘There are students at this school.’ b. *Lesi  sikole si-khona a-bafundi. 7this 7school  7sm-loc.pred  det-students

According to Zeller, -khona is the realization of Pred°, and the subject is its com-plement and is syntactically the predicate, which is interpreted as an argument of existential predication. The default noun class 17 agreement is triggered by an expletive. This can explain the grammaticality of (85a). However, the ungram-maticality of (85b) remains unexplained, as one would expect to be able to intro-duce a locative DP in spec-PredP just as easily as an expletive. After all, that is exactly how Zeller derives cases like (73b) above. Another case is passives. Inter-nal subjects are possible with passives, as in (86), whereas (87) shows (with two different morpheme orderings on the verb) that locative inversion is not possible with a passive verb:

(86) A-bantu a-badala a-ba-shay-w-a. Ku-shay-w-a det-2people  2rel2old  neg-2sm-hit-psv-fs  17sm-hit-psv-fs i-zingane. det-10children ‘Old people aren’t hit. Children are hit.’

(87) a. *Lesi  sikole si-ya-fund-el-w-a. 7this 7school  7sm-dj-study-appl-psv-fs intended: ‘This school is studied at.’ b. *Lesi  sikole si-ya-fund-w-el-a. 7this 7school  7sm-dj-study-psv-appl-fs

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 41: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   463

Thus, the idea that locative inversion, instrumental inversion, and internal sub-jects in expletive subject constructions require a PredP is not enough to explain the distribution of these three constructions.

We will not attempt to explain the distribution of locative and instrumental distribution, but we can provide an explanation of why they cannot occur with non-verbal predication. Locative inversion is ungrammatical for a very simple reason. It can only occur with two types of verbs: with certain unaccusatives (those which take an inherent locative argument) and with verbs of unergative valence or higher supplemented with the applicative suffix. This restriction rules out non-verbal predication without any further apparatus. Instrumental inver-sion is ruled out for a similar reason. Only a semantically limited set of verbs can appear in the construction, as described by Zeller himself. In this way, non-verbal predication is also excluded without recourse to PredP.

6.3 Head movement and the categories v and V

There seems to be a universal constraint or at least a strong universal tendency to prohibit head-movement of adpositions and determiners to PredP. We know of no language where in nominal predication the determiner moves and is conjugated, or where in locative predication the adposition moves and is conjugated. Below is a fictitious example of how such a language would behave:

(88) a. *This  man the-s best cook in the world. this man  the-3sg  best  cook  in  the  world ‘This man is the best cook in the world.’ b. *Johnny  in-s the room. Johnny in-3sg  the  room ‘Johnny is in the room.’

This could be related to the fact that DP’s and PP’s, unlike VP’s, are rarely split up by stranding their head and moving their complement. There are many languages in which the verb is separated from its object complement, such as the many VSO languages (such as the Celtic languages and Classical Arabic), but there are very few languages where the preposition or the determiner is separated from its complement. Conjugated nouns constitute a potential counterexample, because if a language lacks determiners it could be that the noun first undergoes head-movement to D° and then moves further to adjoin to I°. However, in their investi-gation of a few cases of so-called conjugated nouns Matushansky and de Dreu (2010) found that in most cases the “conjugation” could be analyzed as a clitic on

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 42: Subject raising in Zulu and the nature of PredP

464   Leston Chandler Buell and Merijn de Dreu

the noun and that there was no need to assume that any movement had taken place. Thus, it appears that the proposed generalization holds.

7 ConclusionIn the preceding discussion, we examined a syntactic dichotomy in Zulu: verbal and non-verbal predication. The two predication types share certain properties, such as the possibility not only of subject raising, but also of right-dislocation. However, there are characteristics beyond the mere presence or absence of a lexical verb that differentiate these two classes of predication. While subject rais-ing is optional with verbal predicates, with non-verbal predicates (except for -khona) it is obligatory. Furthermore, the morphology makes a clear-cut distinc-tion between verbal and non-verbal predication. This particular partitioning of predicate types is not entirely to be expected. Consider, for example, the familiar characterization of the categories V, N, P, and Adj in terms of [±N] and [±V] fea-tures (Chomsky 1981, p. 48):

(89) Categorial feature matrix

[+N] [−N]

[+V] Adj V

[−V] N P

Assuming such a feature matrix, we might, for example, expect the [+V] predi-cates (those headed by a V or Adj) to pattern differently from the [−V] predicates (P- and Adj-headed predicates). Alternatively, the [+N] predicates could behave differently from the [−N] predicates. What we do not expect is for N-, P-, and Adj-headed predicates to pattern together in opposition to verbal predicates, which is what we actually find.

We have attempted to explain the observed dichotomy by assuming a PredP projection in the case of all non-verbal predicates. While in doing so we are following the lead of Zeller (2010, 2012), our analysis accounts for a different set of facts than his. Both analyses account for the distribution of subject raising, but only ours can account for the morphological alternation. Zeller’s account additionally explains certain supposed semantic properties peculiar to internal subjects. However, we have contested the reality of these supposed properties and have shown that they are essentially the same as those exhibited by internal objects, which would not require the extra functional projection PredP under any analysis.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 43: Subject raising in Zulu and the nature of PredP

Subject raising in Zulu and the nature of PredP   465

Non-verbal predication is sadly still a largely neglected topic in theoretical syntax. Presumably, this is because the most widely studied languages employ a lexically vacuous verb (like be) when no lexical verb is present, thereby masking the dichotomy. However, there are clearly still connections to be made between languages like Zulu and English. Note, for instance, how the existential there con-struction is grammatical with the adjective present but not with young:

(90) a. Therewasaboypresent. b. *Therewasaboyyoung.

These facts seem related to the fact that subject raising in Zulu is obligatory with adjectives but not with -khona, which could be translated as ‘be present’. While many such parallels can be drawn between languages like Zulu and those like English, it remains to be explored which of these are due to structural simi-larities, such as the projection of PredP in both languages, and which are due to semantic factors, such as event structure. We leave these issues for further re-search.

Acknowledgments: We would like to thank our Zulu informant Meritta Xaba, as well as Birgitta Bexten, Ora Matushansky, Jochem Zeller, and two anonymous reviewers. Funding for this research was provided by the NWO (Nederlandse Organisatie voor Wetenschappelijk Onderzoek).

ReferencesBaker, Mark. 2003. Lexical categories: verbs, nouns, and adjectives. Cambridge: Cambridge

University Press.Bowers, John. 1993. The syntax of predication. Linguistic Inquiry 24. 591–656.Buell, Leston Chandler. 2005. Issues in Zulu verbal morphosyntax. Ph.D. dissertation,

University of California, Los Angeles.Buell, Leston Chandler. 2006. The Zulu conjoint/disjoint verb alternation: focus or

constituency? ZAS Papers in Linguistics 43. 9–30.Buell, Leston Chandler. 2007. Semantic and formal locatives: implications for the Bantu locative

inversion typology. SOAS working papers in linguistics: Bantu in Bloomsbury: Special issue on Bantu linguistics. 105–120.

Buell, Leston Chandler. 2008. Predication types and predicate-internal arguments in Zulu. Unpublished manuscript.

Buell, Leston Chandler. 2011. The Nguni augment from a syntactic perspective. Paper presented at the Bantu 4 Conference, Humboldt University, Berlin, April.

Buell, Leston Chandler. 2012. Class 17 as a non-locative noun class in Zulu. Journal of African Languages and Linguistics 33. 1–35.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM

Page 44: Subject raising in Zulu and the nature of PredP

466   Leston Chandler Buell and Merijn de Dreu

Cheng, Lisa & Laura J. Downing. 2009. Where’s the topic in Zulu? The Linguistic Review 26. 207–238.

Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris.Doke, Clement M. 1973 [1927]. Textbook of Zulu grammar. Cape Town: Longman Southern

Africa, sixth edition.Doke, Clement M. & S. Machabe Mokofeng. 1985 [1957]. Textbook of Southern Sotho grammar.

Cape Town: Maskew Miller Longman, second edition.Dreu, Merijn de. 2008. The internal structure of the Zulu DP. Master’s thesis, Leiden University.Dreu, Merijn de & Leston Chandler Buell. 2012. Neuter gender in a sexless language:

The case of Zulu. Linguistics in the Netherlands 29. 41–54.Gowlett, Derek. 2003. Zone S. In Derek Nurse & Gérard Philippson (eds.), The Bantu Languages,

609–638. London: Routledge.Halpert, Claire. 2012. Case, agreement, EPP and information structure: A quadruple-

dissociation in Zulu. In Jaehoon Choi, E. Alan Hogue, Jeffrey Punske, Deniz Tat, Jessamyn Schertz, and Alex Trueman, West Coast Conference on Formal Linguistics (WCCFL) 29.

Hazout, Ilan. 2004. The syntax of existential constructions. Linguistic Inquiry 35. 393–430.Hengeveld, Kees. 1992. Non-verbal predication: theory, typology, diachrony. Berlin: Mouton de

Gruyter.Jordan, Archibald Campbell. 1967. The adjective in the Nguni languages of southern Africa with

special reference to Xhosa. Journal of African Languages 6. 132–145.Matushansky, Ora & Merijn de Dreu. 2010. The myth of conjugated nouns. Paper presented at

the Taalkunde in Nederland-dag (TiN-dag) conference, University of Utrecht, February.Mini, Buyiswa Mavis, Sonwabo Lungile Tshabe, F. M. Shoba, & P. N. Van der Wersthuizen (eds.).

2003. The greater dictionary of isiXhosa, volume 2. IsiXhosa National Lexicography Unit.Muriungi, Peter. 2009. Phrasal movement inside Bantu verbs: deriving affix scope and order in

Kîîtharaka. Tromsø, Norway: University of Tromsø dissertation.Posthumus, Lionel. 2000a. When does ‘become’ BECOME ‘be’? South African Journal of African

Languages 20: 342–352.Posthumus, Lionel. 2000b. The so-called adjective in Zulu. South African Journal of African

Languages 20: 148–158.Poulos, George & Christian T. Msimang. 1998. A linguistic analysis of Zulu. Cape Town:

Via Afrika.Travis, Lisa. 1984. Parameters and the effects of word order variation. Cambridge, MA:

MIT dissertation.Van der Spuy, Andrew. 2001. Grammatical structure and Zulu morphology. Johannesburg, South

Africa, University of the Witwatersrand dissertation.Van der Spuy, Andrew. 1993. Dislocated noun phrases in Nguni. Lingua 90. 335–355.Visser, Marianna. 2008. Definiteness and specificity in the isiXhosa determiner phrase. South

African Journal of African Languages 1. 11–29.Van Eeden, Bernardus Izak Christiaan. 1956. Zoeloe-grammatika. Stellenbosch: Universiteits-

uitgewers.von Staden, P. M. S. 1973. The initial vowel of the noun in Zulu. African Studies 32. 163–181.Zeller, Jochen. 2010. Semantic locative inversion and predication in Zulu. Unpublished

manuscript.Zeller, Jochen. 2012. Instrument inversion in Zulu. Annual Conference on African Linguistics

(ACAL) 42. 134–148.

Brought to you by | Georgetown UniversityAuthenticated | 141.161.91.14

Download Date | 9/27/13 3:27 PM